| name | setup-syv-skills |
| description | One-command onboarding for the Syv.ai skills collection. Checks the environment (git, GitHub CLI auth), seeds a slim CLAUDE.md pointer if absent, confirms which skills resolved, and gives a short orientation. Use when a user has just installed the collection and runs /setup-syv-skills, or asks how to get set up to use the Syv skills. |
| disable-model-invocation | true |
Setup Syv Skills
The onboarding command for the Syv.ai skills collection. After a user installs
the collection (npx skills add syv-ai/agentic-coding-playbook), this readies
the current repo and environment to use it. This is the workshop "pre-flight"
collapsed into one command — run explicitly, never auto-invoked.
This is a prompt-driven skill, not a script. Explore the current state, report
what you found, fix what's missing with the user's go-ahead, and orient them.
Process
1. Check the environment
Run these and report a short PASS/ACTION line for each — don't fix silently:
- Git —
git rev-parse --is-inside-work-tree. If this isn't a repo, offer to git init (ask first).
- GitHub CLI —
gh --version, then gh auth status. The to-prd / to-issues skills default to publishing via gh. If gh is missing or unauthenticated, tell the user to run gh auth login themselves (it's interactive — they should run it, e.g. by typing ! gh auth login). Note that these skills fall back to writing markdown under docs/ if gh stays unavailable, so this is a recommendation, not a blocker.
- Platform note (Windows): on native Windows, prefer a skill's PowerShell variant where one exists (e.g.
guardrails ships both). Any bash-based script a skill installs needs WSL or Git Bash. The brainstorming visual companion is static HTML (no Node, no server) — it just needs a browser and an open/start command.
2. Confirm the skills resolved
List the skills the harness has picked up from this collection so the user can see they're available, grouped the way the collection README groups them. If none resolved, the install didn't register — point them back to the install step and check the plugin is enabled.
3. Seed a slim CLAUDE.md pointer (offer, don't force)
If the repo has no CLAUDE.md, offer to create a slim one. Keep it tiny — only what is both undiscoverable and globally relevant. A good starting point:
# <project name>
<one-line description of what this project is>
<non-standard commands the agent couldn't infer — build/test/run>
Do not run /init or generate a fat CLAUDE.md. Standards belong in a
referenced docs/ file or a skill, not in CLAUDE.md. If a CLAUDE.md already
exists, leave it alone — just mention they can keep it slim.
4. Orient
Close with a two-line orientation: the core loop the collection is built around —
/grill-me → /to-prd → /to-issues → build → review — and that
/setup will wire up the project's dev environment and commit gate if that hasn't
been done yet.
Done when
Related skills
- setup — wire up the project's stack, dev env, and commit gate.
- grill-me — the first step of the core workflow.
